HAL Id: pasteur-03272638https://hal-pasteur.archives-ouvertes.fr/pasteur-03272638v2

Preprint submitted on 6 Sep 2021

HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, estdestinée au dépôt et à la diffusion de documentsscientifiques de niveau recherche, publiés ou non,émanant des établissements d’enseignement et derecherche français ou étrangers, des laboratoirespublics ou privés.

Distributed under a Creative Commons Attribution - NonCommercial - NoDerivatives| 4.0International License

Epidemiology and control of SARS-CoV-2 epidemics inpartially vaccinated populations: a modeling study

applied to FrancePaolo Bosetti, Cécile Tran Kiem, Alessio Andronico, Vittoria Colizza, Yazdan

Yazdanpanah, Arnaud Fontanet, Daniel Benamouzig, Simon Cauchemez

To cite this version:Paolo Bosetti, Cécile Tran Kiem, Alessio Andronico, Vittoria Colizza, Yazdan Yazdanpanah, et al..Epidemiology and control of SARS-CoV-2 epidemics in partially vaccinated populations: a modelingstudy applied to France. 2021. �pasteur-03272638v2�

Epidemiology and control of SARS-CoV-2 epidemics in partiallyvaccinated populations: a modeling study applied to France

Paolo Bosetti1,*, Cécile Tran Kiem1,2,*, Alessio Andronico1, Vittoria Colizza3,Yazdan Yazdanpanah4,5, Arnaud Fontanet6,7, Daniel Benamouzig8, SimonCauchemez1

Affiliations:

1. Mathematical Modelling of Infectious Diseases Unit, Institut Pasteur,UMR2000, CNRS, Paris, France

2. Collège Doctoral, Sorbonne Université, Paris, France3. INSERM, Sorbonne Université, Institut Pierre Louis d’Epidémiologie et de

Santé Publique, Paris, France4. Université of Paris, INSERM UMR 1137 IAME, Paris, France.5. Department of Infectious Diseases, Assistance Publique-Hôpitaux de Paris,

Bichat–Claude-Bernard University Hospital, Paris, France.6. Emerging Diseases Epidemiology Unit, Institut Pasteur, Paris, France7. PACRI Unit, Conservatoire National des Arts et Métiers, Paris, France8. Sciences Po - Centre de sociologie des organisations and Chaire santé -

CNRS, Paris, France

*Contributed equally

Corresponding author:

Simon Cauchemez

Mathematical Modelling of Infectious Diseases Unit

Institut Pasteur,

28 rue du Dr Roux,

75015, Paris, France

simon.cauchemez@pasteur.fr

1

Abstract

Vaccination will change SARS-CoV-2 epidemiology. We used an age stratified

compartmental model calibrated to French data to anticipate these changes and determine

implications for epidemic control, assuming vaccines reduce the risk of hospitalisation,

infection and transmission if infected by 95%, 60% and 50%, respectively. In our baseline

scenario (R0=5; vaccine coverage of 70%-80%-90% among 12-17, 18-59 and ≥60 y.o.),

important stress on healthcare is expected without measures. Unvaccinated adults ≥60 y.o.

represent 3% of the population but 43% of hospitalisations. Children aged 0-17 y.o.

represent a third of infections and are responsible for almost half of transmissions.

Unvaccinated individuals have a disproportionate contribution to transmission so that

measures targeting them may help maximize epidemic control while minimizing costs for

society compared to non-targeted approaches. With the Delta variant, vaccinated individuals

are well protected against hospitalization but remain at risk of infection and should therefore

apply protective behaviours (e.g. mask wearing). Control strategies should account for the

changing SARS-CoV-2 epidemiology.

2

Introduction

The SARS-CoV-2 pandemic that started in December 2019 has caused more than 3.8

million deaths around the world and led healthcare systems at the brink of collapse in many

countries. In addition, the drastic control measures that were implemented to limit its impact

have had dramatic socio-economic consequences.

Vaccines have proved effective at reducing the severity of SARS-CoV-2 infection,1 the risk of

infection2 and transmission3. A number of modelling studies evaluated how vaccination will

help mitigate a SARS-CoV-2 epidemic rebound this autumn, highlighting that it might be

difficult to fully relax control measures given the high transmissibility and severity of

SARS-CoV-24–7. These studies assessed impact on key health metrics (e.g. number of

hospitalisations and death) and identified the level of social distancing that would remain

necessary as a function of vaccine coverage. A question that has received less attention is

that, in this new era where a large part of the population is vaccinated, the epidemiology of

SARS-CoV-2 (Who is infected? Who transmits? Who is hospitalized?) will be different from

what it was prior to the distribution of vaccines8. It is important to anticipate these changes to

determine how control measures might evolve to ensure they maintain the epidemic under

control while minimizing costs for society. For example, the expectation that unvaccinated

individuals will have a higher contribution to infections, transmissions and hospitalisations

has led some countries to introduce control strategies specifically targeting this population.

This is the case of France: confronted to a rapid rise in Delta cases and a plateau in

vaccinations in June-July 2021, French authorities announced in July that a sanitary pass,

i.e. a proof of completed vaccination, recent infection or recent negative test, would be

required to access places such as bars, restaurants and cinemas. The announcement led to

an important surge in vaccination appointments and in vaccine coverage. A number of

European countries introduced similar measures.

3

Here, we developed a mathematical model to characterize the epidemiology of SARS-CoV-2

in a partially vaccinated population and evaluate in this new context the contribution to

transmission and healthcare burden of individuals of different ages and vaccination status.

This information is used to ascertain different control strategies, including repeated testing

and non-pharmaceutical measures, targeting the whole population or subgroups such as

unvaccinated individuals to optimally mitigate an autumn epidemic rebound. This is also an

opportunity to revisit impact assessment accounting for the increased transmissibility and

severity of the Delta variant as well as the reduction in vaccine protection against infection

associated with this variant. We consider Metropolitan France as a case study.

Results

Baseline scenario and no control measures

We first present results under the assumption that control measures are completely relaxed

in the autumn 2021, for our baseline scenario (basic reproduction number R0=5 and a

vaccine coverage of 70%-80%-90% among teenagers, adults aged 18-59 years old (y.o.)

and over 60, respectively). In this case, our model anticipates a wave characterized by a

peak of 5,200 hospital admissions per day which is larger than the peaks observed in France

during the two pandemic waves of 2020.

We anticipate that the roll-out of vaccines will modify the epidemiology of SARS-CoV-2. In a

context where most adults are vaccinated but vaccine coverage remains limited among

children (0-17 y.o.), we expect 33% of infections will occur in this age group, even though

they only represent 22% of the population and are assumed to be less susceptible to

SARS-CoV-2 infection than adults (Figure 1A). In each age group, unvaccinated individuals

are overrepresented among infected people while vaccinated individuals are

under-represented (Figure 1B). For example, the risk of infection for an unvaccinated

4

individual is RR=1.9 times higher than that of a vaccinated individual among those aged

18-59 y.o (RR=1.3 among 0-17 y.o. and RR=2.1 among over 60; Supplementary Table 1).

Overall, unvaccinated individuals represent 29% of the population but 46% of infections.

Their contribution to the transmission process is even higher with a risk of transmission from

an unvaccinated individual that is 4.3 times higher than that from a vaccinated individual

(Figure 1C-D).

Vaccination will also impact the age distribution of those hospitalised. While 74% of

hospitalisations occurred among those older than 60 y.o. in the pre-vaccination era, this

proportion is expected to drop to 65% in our baseline scenario. In parallel, the proportion of

18-59 y.o. among hospitalized individuals increases from 25% in the pre-vaccination era to

30% (Figure 1E). The small group of unvaccinated adults that are older than 60 y.o. has a

disproportionate impact on the stress to the healthcare system. They represent 10% of their

age group but 66% of hospitalisations from that age group (RR: 17.2); and 3% of the general

population but 43% of all hospitalisations (RR: 26.7) (Figure 1F). Even though we assume

that the vaccine is 95% effective against the risk of hospitalisation, in a context where

vaccine coverage is high among older individuals, 28% of hospitalisations occur among

vaccinated people (Figure 1F).

Baseline scenario with control measures

We then investigate the impact of different control strategies targeting different groups for

our baseline scenario with a vaccination coverage 70%-80%-90% and R0=5. Weekly testing

of 50% of unvaccinated individuals aged ≥12 y.o. could reduce the peak of hospitalisations

by 19% (range: 16-23% for 20-30% of the population infected prior to September 1st 2021) if

an autotest is used and 22% (19-27%) if the test is performed by a professional (Figure 2A).

In contrast, if the same number of tests were distributed randomly among individuals aged

≥12 y.o. irrespective of vaccination status, the reductions in hospital admissions would only

5

be of 8% (7%-10%) and 10% (8%-12%), respectively. The reduction in the peak of

hospitalisations would be much larger if 50% of unvaccinated individuals aged ≥12 y.o.

agreed to get vaccinated instead of being repeatedly tested (68% vs 19%; Figure 2A), for a

cost that would be 4.5 times lower (0.16 vs 0.72 billion euros; Supplementary Figure S2).

Moreover, only vaccination would be able to reduce the peak of hospital admissions below

the peaks observed in the 2020 spring and fall waves.

Non-pharmaceutical interventions applied to all and reducing the overall transmission rates

by 10%, 20%, 30% and 40% would reduce the peak of hospitalisations by 39%, 70%, 90%

and 97%, respectively (Figure 2B). If these measures were only targeted towards the few

unvaccinated individuals (29% of the population), we could still reach 27%, 51%, 72% and

87% peak reductions, respectively. In both scenarios, reducing transmission rates by 20%

would be sufficient to make the peak of hospitalisations drop below the levels observed

during the second wave of 2020. Given the residual risk of infection in vaccinated individuals

with the Delta variant, substantial gains can be made if vaccinated individuals mitigate their

risk of infection for example by wearing masks.

Sensitivity analyses

Keeping in mind that uncertainties remain about R0 and the vaccine coverage in the Autumn,

we investigate how our results change if we depart from our baseline assumptions. Figure 3

shows the expected size of the Autumn peak in hospital admissions, for different values of

R0 and vaccine coverages, considering scenarios where control measures can target

unvaccinated individuals only or the whole population leading to reductions of transmission

rates in the targeted population between 0% (no control) and 40%. As expected the size of

the peak increases with R0 and declines with the vaccination coverage. For R0=3, which was

the value estimated for the historical lineage, we would not anticipate an epidemic rebound

for the vaccine coverage expected to be achieved in France. For R0=4, the peak is expected

to be below the ones of 2020 even if measures are fully relaxed. For R0=5, in the absence of

6

control measures, the peak could remain below the one of the Autumn 2020, if vaccine

coverage was increased to 70% in teenagers, 90% in 18-59 y.o. and 95% in those over 60

y.o.. For R0=6, increasing the vaccine coverage to these levels would still not allow a full

relaxation of control measures and the implementation of control measures would be

necessary to further mitigate impact on healthcare. Overall, high vaccination coverage and

even limited control of transmission can help mitigate an epidemic rebound. The age

distribution of infected and hospitalized individuals depends on vaccine coverage in the

different age groups (Figure 4, Figure S3-S4). For example, when vaccine coverage in those

over 60 y.o. increases from 90% to 95%, the contribution of this age group to hospitalisations

drops from 65% to 55%. Those distributions are relatively robust to a change in R0

(Supplementary Figure S3-S4).

Comparing our baseline scenario (60% reduction in the risk of infection given Delta) to that

with an 80% reduction in the risk of infection that was considered prior to the rise of Delta,

we find that the lower protection conferred by vaccines against Delta infection substantially

degrades projections, with a peak of hospitalisations that roughly doubles when moving from

80% to 60% protection (Figure 3). This reduction of protection against infection also

increases the contribution of vaccinated individuals to infections: they represent 34% of

infections with a protection of 80% compared to 54% with a protection of 60% (Figure 4,

Figure S6). As a consequence, compared to vaccinated individuals, unvaccinated individuals

are 4 and 10 times more likely to transmit in scenarios with a protection against infection of

60% and 80%, respectively, for R0=5 and a vaccine coverage of 70%-80%-90%.

For 90% protection against hospitalisation instead of 95% in our baseline scenario, we

expect a 28% increase in the peak of hospitalisations (Figure 3). This also increases the

proportion of vaccinated individuals among those hospitalized from 28% to 44% (Figure 4 ,

Figure S7).

7

Comparing our baseline scenario (70% of teenagers vaccinated) to one where teenagers are

not vaccinated (Supplementary Figure S8), our results suggest that the vaccination of

teenagers may substantially reduce the stress on the healthcare system. For example, if

80% of 18-59 y.o. and 90% of over 60 are vaccinated, the vaccination of 70% of teenagers

could reduce the peak of hospitalisations by 66% and 40% for R0=4 and 5, respectively,

compared to a scenario where they are not vaccinated.

If children aged 0-9 y.o. are 50% less infectious than adults in addition to being 50% less

susceptible, the proportion of children among infections decreases from 33% to 31% while

the proportion among those that cause infection drops from 43% to 36% (Supplementary

Figure S5).

Discussion

Countries with partially vaccinated populations enter a new era in the control of the

SARS-CoV-2 epidemic. However, given the high transmissibility and severity of the Delta

variant and the reduced efficacy of vaccines against infection by this variant, SARS-CoV-2

may continue to generate substantial stress on healthcare in the absence of mitigation

measures, even with high vaccine coverage. Nonetheless, the partial vaccination of the

population modifies the epidemiology of SARS-CoV-2. Here, we used a mathematical model

applied to Metropolitan France to anticipate these changes and determine how control

measures might evolve in the autumn 2021 to maximize their impact while minimizing costs.

This autumn, the stress on the healthcare system in the absence of any control measures

will depend on the vaccine coverage and the transmission potential R0 of the dominant

variant. R0 was around 3 for the historical lineages9. The Alpha variant that is currently

dominant in France was found to be about 50% more transmissible than historical

lineages10–12 and the Delta variant that is now dominant might be more than 50% more

transmissible than the Alpha variant13. If we simply apply these multiplicative terms, R0 might

be as high as 7 for the Delta variant. However, it is possible that transmissibility differences

8

between variants change with control conditions. We therefore considered R0=5 in our

baseline analysis and explored values between 3 and 6 in our sensitivity analyses. For R0≥5

which appears likely for the Delta variant and under our baseline vaccine coverage of

70%-80%-90% among teenagers, younger and older adults, we anticipate an important

stress on the healthcare system in the absence of any control measure (Figure 3A). Ongoing

efforts to control transmission should therefore be maintained. On a more positive note,

thanks to vaccination, the intensity of control measures necessary to maintain

hospitalisations at manageable levels should be substantially less than what was required

before the roll-out of vaccines. Indeed, while lockdowns used in 2020 reduced transmission

rates by 70-80%9, we find that reductions of the order of 20-30% might now be sufficient.

Such reductions might potentially be achieved through protective measures (e.g. masks,

hand hygiene), a certain degree of social distancing, the sanitary pass and

Test-Trace-Isolate.

Since vaccines reduce the risk of infection and of transmission if infected, our model

anticipates that unvaccinated individuals will contribute more to disease spread than

vaccinated ones. Since vaccine coverage among children aged 0-17 y.o. will be low relative

to that in adults, we anticipate a strong increase of children’s contribution, with about one

third of infections occurring in children and 43% being due to this group in our baseline

scenario. Adults that are not vaccinated will also disproportionately contribute to the stress

on the healthcare system. This is particularly true for those that are older than 60 y.o. In our

baseline scenario, this group represents 3% of the population but 43% of hospital

admissions.

These observations have important implications for epidemic control. First, they show the

importance of obtaining near perfect vaccine coverages in older age groups that contribute

disproportionately to the stress on the healthcare system. This likely requires the

development of strategies where authorities reach out to individuals to facilitate their access

to vaccines. Second, we anticipate that, in a population that is partially vaccinated, gains

9

achieved thanks to social distancing measures are larger when reducing the contacts of

unvaccinated individuals rather than those of vaccinated ones. This suggests that, in this

new era, control measures targeting unvaccinated individuals (for example with the use of

the sanitary pass) may help maximizing epidemic control. Such a targeting strategy also

raises ethical and social issues. From an economic perspective, targeting unvaccinated

individuals maximizes the effectiveness of control while minimizing the cost to society. This is

consistent with the theory that in situations where a small group of individuals contributes

disproportionately to the spread of disease, it is optimal to target that group. However,

targeting the unvaccinated leads to forms of discrimination, more or less severely felt. While

it is true that discrimination between the vaccinated and unvaccinated is to some extent the

result of individual choices, as vaccines are widely available, these choices are nevertheless

socially stratified and correlated with age and socioeconomic status. Moreover, the

restrictions put in place are not chosen by individuals but defined by the authorities. Choices

may therefore be perceived as discrimination, especially by the unvaccinated. in France,

while the sanitary pass has been widely accepted, it has also actively mobilized against it

minority segments of the population.

Recent data indicate a reduction in vaccine efficacy against infection by the Delta variant

and a waning of immunity against infection, with protection against hospitalisation remaining

elevated. These changes have important implications for the management of the epidemic

since we expect they will facilitate viral circulation even in highly vaccinated populations and

eventually increase the stress on the healthcare system. This means that, compared to the

first half of 2021, there is a higher risk that vaccinated individuals get infected and transmit

the virus. As a consequence, measures reducing the risk of infection and transmission such

as the wearing of the mask should apply to vaccinated individuals in situations where

transmission is possible (e.g. indoors).

The situation of children is a particular source of concern. Children aged <12 y.o. do not

have access to vaccines yet and vaccine coverage remains lower among teenagers due to

10

later access to the vaccine. While children mostly develop mild SARS-CoV-2 infections, it is

essential to secure their access to education and a normal social life and to protect their

mental health. Low vaccine coverage among children puts them at risk of being exposed to

class closures, with a deleterious impact on their education and mental health14. The

vaccination of children would insulate them from that risk. In the case of children, the ethical

and social problems are exacerbated. On the vaccination side, discrimination arises from the

fact that children cannot be seen as making voluntary choices between vaccination and

social restrictions. Vaccination is not yet offered under the age of 12, and beyond the age of

12, the "choice" to be vaccinated depends primarily on the family environment. As for other

measures potentially targeted at schools, a wide range of instruments is available (from

mask wearing to physical distancing, air filtration, iterative self-testing, closing rules,

dedicated tracing, isolation of family members...) and could help mitigate impact15 but their

targeted implementation would disproportionately affect young people and their families,

raising questions of social justice if society at large is less directly targeted, particularly in

certain age groups.

This assessment is performed in a context of uncertainty about the value of R0 and vaccine

coverage in the autumn. Our model makes a number of simplifying assumptions. We ignore

a potential decay of immunity, whether immunity was acquired through natural infection or

vaccination, however we account for the reduced vaccine efficacy against infection

estimated for the Delta variant. We consider a national model for France and do not account

for spatial heterogeneities, that are important16. We considered a 'leaky' vaccine that exhibits

failure in degree, as most SARS-CoV-2 models5,7,17,18. This assumption could lead to larger

epidemic sizes than models with 'all-or-nothing' vaccines. For this reason, and given the

uncertainty on the basic reproductive ratio for Delta, we performed a sensitivity analysis on

R0.

11

We used a mathematical model to anticipate how the epidemiology of SARS-CoV-2 may

change in partially vaccinated populations and investigate implications for the control of a

possible epidemic rebound this autumn.

Methods

Deterministic model

We developed a deterministic age-stratified compartmental model describing the spread of

SARS-CoV-2 in metropolitan France. The model, which accounts for French age-specific

contact patterns19, has been described in detail elsewhere.9 It accounts for a gradient of

severity with age20, assuming that Delta VOC is 50% more severe than Alpha VOC21 while

Alpha VOC is 40% more severe than previously circulating strains22. It has been extended to

account for the roll-out of vaccines4 as well as the deployment of self-administered rapid

antigenic tests.23 A full description of the model and equations is reported in the Supplement.

Scenarios

Vaccine coverages and characteristics

Considering the Delta variant, we assume that vaccines are 95% effective at reducing the

risk of hospitalisation1, 60% at reducing the risk of infection24 (impact on susceptibility) and

50% at reducing the infectivity of vaccinated individuals3. In a sensitivity analysis, we show

results if vaccines are 80% effective at reducing the risk of infection2 (which was the scenario

considered prior to the rise of Delta) and 90% effective against hospitalisation. We build

several scenarios regarding vaccine coverage achieved in the different age groups by

September 1st, 2021: 90% or 95% among those older than 60 years old (y.o.); 60%, 80% or

90% among those aged 18-59 y.o. and 0%, 30% or 70% among the 12-17 y.o. (called

teenagers in the following). To give some context, 89% of those older than 60 y.o., 84% of

the 18-59 y.o. and 61% of the 12-17 y.o. have received a first dose of vaccines against

SARS-CoV-2 by August 25th, 2021. In our baseline scenario that we label 70%-80%-90%,

12

we assume vaccination coverage will reach 70%, 80% and 90% among 12-17 y.o., 18-59

y.o. and over 60 on September 1st, 2021. In this analysis, we consider that the vaccine

coverage corresponds to the proportion of the population having acquired vaccine protection

after two doses if required.

Epidemic dynamics with and without control measures

We assume that, by September 1st, 2021, 25% (range: 20-30%) of the French population

has been infected by SARS-CoV-2, benefiting from natural protection against reinfection. We

then explore scenarios where different types of control measures are implemented.

First, we explore scenarios where control measures are completely relaxed in the Autumn.

These scenarios are characterized by the basic reproduction number R0, i.e. the average

number of persons infected by a case in a population with no immunity and no control

measures. In March 2020, R0 was estimated around 3 in France prior to the implementation

of a nation-wide lockdown.9 The emergence of more transmissible variants of concerns

(VOC) (such as the Alpha and Delta VOCs)11–13,25 is expected to increase R0. We therefore

explore scenarios in which R0 ranges between 3.0 and 6.0 when measures are completely

relaxed, considering R0=5 as our baseline scenario. We assume that from September 1st,

2022, the structure of contacts in the population comes back to the one measured during the

pre-pandemic period.19

We then consider scenarios where different types of control measures are implemented,

targeting different groups:

● Iterative testing: we assume that a proportion of the population is targeted for iterative

testing with antigenic tests. These individuals test at regular intervals (every 7 days in

the baseline scenario; twice a week and every 2 weeks in sensitivity analyses). We

assume that individuals testing positive isolate in a way that reduces onward

transmission by 75%. We consider a scenario where 50% of unvaccinated individuals

aged ≥12 y.o. get tested iteratively and a scenario where the same number of

13

individuals randomly drawn among individuals aged ≥12 y.o. (vaccinated or

unvaccinated) are tested iteratively. We consider scenarios where the antigenic test is

performed by the individual (self-swabbing and reading of the result; sensitivity: 75%) or

by a professional (sensitivity 90%). In a sensitivity analysis, we also explore a scenario

where 25% of unvaccinated individuals ≥12 y.o. get tested iteratively.

● Non-pharmaceutical interventions: Non-pharmaceutical interventions such as social

distancing, protective measures and mask wearing may be used to reduce transmission

rates. We consider scenarios where such measures target the whole population, leading

to reductions of transmission rates of 10%, 20%, 30% or 40% from any infected

individual, whether they have been vaccinated or not. We also consider scenarios

where such measures only target unvaccinated individuals, leading to reductions of

transmission rates of 10%, 20%, 30% or 40% from unvaccinated individuals, while

transmission rates from vaccinated individuals remain unchanged.

● Increased vaccine coverage among unvaccinated individuals: We compare the

performance of these interventions to that obtained if 50% of the unvaccinated

individuals aged ≥12 y.o. were to get vaccinated.

Children are defined as individuals aged 0-17 y.o. We assume that children aged 0-9 y.o. are

50% less susceptible to infection than adults while those aged 10-17 y.o. are 25% less

susceptible to infection than adults 9,26. In a sensitivity analysis, we also assume that children

aged 0-9 y.o. are 50% less infectious than adults.

We assume an antigenic test costs 5 euros if performed by the individual, 11 euros if

performed by a professional and a 2-doses vaccine costs 32 euros. Models are run until

March 20th 2022.

Funding: We acknowledge financial support from the Investissement d’Avenir program, the

Laboratoire d’Excellence Integrative Biology of Emerging Infectious Diseases program (grant

14

ANR-10-LABX-62-IBEID), HAS, Santé Publique France, the INCEPTION project

(PIA/ANR-16-CONV-0005), the European Union’s Horizon 2020 research and innovation

program under grant 101003589 (RECOVER) and 874735 (VEO), AXA and Groupama.

Author contributions: PB, CTK and SC conceived the study. PB, CTK and AA performed

the analyses. PB, CTK and SC wrote the first draft. All authors contributed to revisions of the

manuscript.

Competing interests: None declared.

Data and materials availability: Data and code will be published online upon publication.

References1. Dagan, N. et al. BNT162b2 mRNA Covid-19 Vaccine in a Nationwide Mass Vaccination

Setting. N. Engl. J. Med. 384, 1412–1423 (2021).

2. Hall, V. J. et al. COVID-19 vaccine coverage in health-care workers in England and

effectiveness of BNT162b2 mRNA vaccine against infection (SIREN): a prospective,

multicentre, cohort study. Lancet 397, 1725–1735 (2021).

3. Harris, R. J. et al. Impact of vaccination on household transmission of SARS-COV-2 in

England. (2021).

4. Tran Kiem, C. et al. Short and medium-term challenges for COVID-19 vaccination: from

prioritisation to the relaxation of measures. (2021).

5. Moore, S., Hill, E. M., Tildesley, M. J., Dyson, L. & Keeling, M. J. Vaccination and

non-pharmaceutical interventions for COVID-19: a mathematical modelling study.

Lancet Infect. Dis. 21, 793–802 (2021).

6. Viana, J. et al. Controlling the pandemic during the SARS-CoV-2 vaccination rollout.

Nat. Commun. 12, 3674 (2021).

7. Leung, K., Wu, J. T. & Leung, G. M. Effects of adjusting public health, travel, and social

measures during the roll-out of COVID-19 vaccination: a modelling study. Lancet Public

Health 6, e674–e682 (2021).

15

8. Galmiche, S. et al. Exposures associated with SARS-CoV-2 infection in France: A

nationwide online case-control study. The Lancet Regional Health - Europe vol. 7

100148 (2021).

9. Salje, H. et al. Estimating the burden of SARS-CoV-2 in France. Science 369, 208–211

(2020).

10. Volz, E. et al. Assessing transmissibility of SARS-CoV-2 lineage B.1.1.7 in England.

Nature 593, 266–269 (2021).

11. Gaymard, A. et al. Early assessment of diffusion and possible expansion of

SARS-CoV-2 Lineage 20I/501Y.V1 (B.1.1.7, variant of concern 202012/01) in France,

January to March 2021. Euro Surveill. 26, (2021).

12. Davies, N. G. et al. Estimated transmissibility and impact of SARS-CoV-2 lineage

B.1.1.7 in England. Science 372, (2021).

13. Campbell, F. et al. Increased transmissibility and global spread of SARS-CoV-2 variants

of concern as at June 2021. Eurosurveillance vol. 26 (2021).

14. YoungMinds. Coronavirus: Impact on young people with mental health needs. Survey 4:

February 2021. https://youngminds.org.uk/media/4350/coronavirus-report-winter.pdf

(2021).

15. Colosi, E. et al. Self-testing and vaccination against COVID-19 to minimize school

closure. medRxiv 2021.08.15.21261243 (2021).

16. Hozé, N. et al. Monitoring the proportion of the population infected by SARS-CoV-2

using age-stratified hospitalisation and serological data: a modelling study. Lancet

Public Health 6, e408–e415 (2021).

17. Hogan, A. B. et al. Within-country age-based prioritisation, global allocation, and public

health impact of a vaccine against SARS-CoV-2: A mathematical modelling analysis.

Vaccine 39, 2995–3006 (2021).

18. Matrajt, L., Eaton, J., Leung, T. & Brown, E. R. Vaccine optimization for COVID-19: Who

to vaccinate first? Sci Adv 7, (2020).

19. Béraud, G. et al. The French Connection: The First Large Population-Based Contact

16

Survey in France Relevant for the Spread of Infectious Diseases. PLoS One 10,

e0133203 (2015).

20. Lapidus, N. et al. Do not neglect SARS-CoV-2 hospitalization and fatality risks in the

middle-aged adult population. Infectious Diseases Now vol. 51 380–382 (2021).

21. Twohig, K. A. et al. Hospital admission and emergency care attendance risk for

SARS-CoV-2 delta (B.1.617.2) compared with alpha (B.1.1.7) variants of concern: a

cohort study. Lancet Infect. Dis. (2021) doi:10.1016/S1473-3099(21)00475-8.

22. Bager, P. et al. Increased Risk of Hospitalisation Associated with Infection with

SARS-CoV-2 Lineage B.1.1.7 in Denmark. SSRN Electronic Journal (2021)

doi:10.2139/ssrn.3792894.

23. Bosetti, P. et al. Impact of mass testing during an epidemic rebound of SARS-CoV-2: a

modelling study using the example of France. Euro Surveill. 26, (2021).

24. Tartof, S. Y. et al. Six-Month Effectiveness of BNT162B2 mRNA COVID-19 Vaccine in a

Large US Integrated Health System: A Retrospective Cohort Study. (2021)

doi:10.2139/ssrn.3909743.

25. Public Health England. Investigation of SARS-CoV-2 variants of concern: variant risk

assessments - Risk assessment for SARS-CoV-2 variant: Delta (VOC-21APR-02,

B.1.617.2).

https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment

_data/file/992981/10_June_2021_Risk_assessment_for_SARS-CoV-2_variant_DELTA.

pdf (2021).

26. Viner, R. M. et al. Susceptibility to SARS-CoV-2 Infection Among Children and

Adolescents Compared With Adults: A Systematic Review and Meta-analysis. JAMA

Pediatr. 175, 143–156 (2021).

17

Figure 1: Contribution of groups defined by their age and vaccination status toinfections, disease spread and hospital burden, in our baseline scenario with R0=5and a vaccine coverage of 70%-80%-90% among 12-17 y.o., 18-59 y.o. and over 60 y.o.Age distribution of new infections A. in the entire population and B. among vaccinated andunvaccinated individuals. Proportion of infections C. attributable to different age groups andD. attributable to different age groups among vaccinated and unvaccinated individuals. Agedistribution of hospitalisations E. in the entire population and F. among vaccinated and

18

unvaccinated individuals. In all panels, the diamonds indicate the age distribution of thedifferent groups in the population.

19

Figure 2: Comparison of the impact of control strategies targeting the entirepopulation vs unvaccinated individuals only, in our baseline scenario with R0=5 and avaccine coverage of 70%-80%-90% among 12-17 y.o., 18-59 y.o. and over 60 y.o. A.Peak in daily hospital admissions under different testing strategies. Baseline - nointervention; Autotest unvaccinated - 50% of the unvaccinated individuals aged ≥12 y.o. aretested weekly (sensitivity of 75%); Autotest random - the same number of individuals as inthe Autotest unvaccinated are tested but among individuals aged ≥12 y.o., irrespective ofvaccine status; Antigenic unvaccinated - same as in Autotest unvaccinated but with tests

20

performed by a professional (sensitivity of 90%); Antigenic random - same as in Autotestrandom but with tests performed by a professional (sensitivity of 90%); Vaccinate - 50% ofthe unvaccinated individuals aged ≥12 y.o. are vaccinated. B. Peak in daily hospitaladmissions under non-pharmaceutical interventions of varying intensities. Baseline - nointervention; Reduction of x% unvaccinated - The transmission rate of unvaccinatedindividuals is reduced by x%; Reduction of x% all - The transmission rate at the populationlevel is reduced by x%. We assume 25% of the population has acquired protection throughnatural infection (range 20%-30% corresponding to the vertical bars).

21

Figure 3: Expected size of the peak of hospitalisations when non-pharmaceuticalinterventions target unvaccinated individuals only or the whole population, as afunction of the basic reproduction number R0, vaccine coverage in the 12-17 y.o.,18-59 y.o. and over 60 y.o. and for different efficacy of the vaccine against the risk ofinfection or hospitalisation. Non-pharmaceutical interventions reduce the transmissionrate of unvaccinated individuals (points) or the whole population (triangles) by 0%, 10%,20%, 30%, 40%. R0 takes the values 3.0 (top row, A-B-C), 4.0 (D-E-F), 5.0 (G-H-I), and 6.0(bottom row, J-K-L). In the baseline scenario (left column) we assume that the vaccines are95% effective at reducing the risk of hospitalisation, 60% at reducing the risk of infectionand 50% at reducing the infectivity of vaccinated individuals. In sensitivity analyses, weconsider an 80% reduction against infection (middle column) and 90% reduction against

22

hospitalisation (right column). We assume 25% of the population has acquired protectionthrough natural infection (range 20%-30% corresponding to the vertical bars). Horizontallines indicate the peak of daily hospital admissions observed during the first (dashed line)and the second (dotted line) epidemic wave of 2020.

23

Figure 4: Proportion of infections (A,C,E) and hospitalizations (B,D,F) among groupsdefined by their age and vaccination status as a function of the vaccine coverage inthe 12-17 y.o., 18-59 y.o. and over 60 y.o.. In the baseline scenario (A-B) we assume thatvaccines are 95% effective at reducing the risk of hospitalisation, 60% at reducing the risk ofinfection and 50% at reducing the infectivity of vaccinated individuals. In (C-D), we assumea vaccine efficacy at reducing the risk of infection of 80%. In (E-F), we assume a vaccineefficacy at reducing the risk of hospitalisation of 90%. The distribution is reported forinfections and hospitalizations occurring between September 1st, 2021 and March 20th,

24

2022 (end of the study period), for R0=5.0. We assume 25% of the population has acquiredprotection through natural infection.

25

Supplement for: Epidemiology and control of SARS-CoV-2epidemics in partially vaccinated populations: a modeling studyapplied to France

Supplementary materials

Model parametrization

We developed a deterministic SEEIR model stratified by age similar to the one used in Saljeet al.1 The model has been extended to account for the roll-out of vaccines2 as well as thedeployment of self-administered rapid antigenic tests3. The metropolitan French population isdivided into the following 13 age groups: [0-10), [10-18), [18-30), [30-40), [40-45), [45-50),[50-55), [55-60), [60-65), [65-70), [70-75), [75-80) and ≥ 80. We assume that individualsaged 0-9 y.o. and 10-17 y.o are respectively 50% and 25% less susceptible compared toadults4,5. The model is implemented with the R software using the odin package6.

Transmission model accounting for iterative testing

Upon infection, susceptible individuals (S) move to the compartment E1. After an averageduration of 4.0 days, infected individuals move to the E2 compartment where they becomeinfectious. They stay in this compartment for an average duration of 1.0 day before movingto the I compartment (IM for mild infections or IH for infections requiring an admission inhospital) in which a fraction of them will develop symptoms. The average length of stay in I isequal to 3.0 days. The proportion of infected people who will need to be admitted to thehospital is age-dependent and accounts for the increased severity associated with the Deltavariant. Specifically, Delta VOC is assumed to be 50% more severe than Alpha VOC7,whereas Alpha VOC is assumed to be 40% more severe than historical strains8. Theage-dependent probability of hospitalization for the historical strains is obtained from Lapiduset al9. Finally, individuals in the IM compartment will recover (R compartment), whileindividuals in the IH will move to the ĪH compartment before being admitted in hospital(compartment H). Individuals who have been vaccinated follow the same path as those whohave not been vaccinated, but they are less susceptible to infection, have a reduced risk ofbeing hospitalized, and are less likely to transmit the disease 2.

Our framework accounts for the deployment of iterative testing strategies. Upon receiving apositive test, we assume that infectious individuals (in compartments E2, IM, and IH)detected isolate, resulting in a reduction of their transmission rate by 75%. This correspondsto the compartments E2iso, IMiso, and IHiso. We assume that individuals tested while in E1remain undetected and therefore do not isolate. The average length of stay in isolatedcompartments is identical to the one in non-isolated ones.

Model Equations

The model can be described by the following set of ordinary differential equations:

1

𝑑𝑆𝑖/𝑑𝑡 = − 𝑆

𝑖 β

( Σ𝑗 𝐶

𝑖𝑗 ((1 − ρ

𝑖𝑛𝑡)(𝐸2

𝑗+ 𝐼

𝑗𝑚𝑖𝑙𝑑 + 𝐼

𝑗ℎ𝑜𝑠𝑝 + (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜

𝑗 + 𝐼𝑖𝑠𝑜

𝑗𝑚𝑖𝑙𝑑 + 𝐼𝑖𝑠𝑜

𝑗ℎ𝑜𝑠𝑝) )

+ (1 − ρ𝑣𝑖𝑛𝑡

)(1 − 𝑉𝐸𝑖𝑛𝑓

)(𝐸2𝑣𝑗 + 𝐼𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑣

𝑗

ℎ𝑜𝑠𝑝+ (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜𝑣

𝑗 + 𝐼𝑖𝑠𝑜𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑖𝑠𝑜𝑣

𝑗

ℎ𝑜𝑠𝑝)))/𝑁

𝑗)

𝑑𝑆𝑣𝑖/𝑑𝑡 =

− (1 − 𝑉𝐸𝑠𝑢𝑠𝑐

) 𝑆𝑣

𝑖 β( Σ

𝑗 𝐶

𝑖𝑗 ((1 − ρ

𝑖𝑛𝑡)(𝐸2

𝑗+ 𝐼

𝑗𝑚𝑖𝑙𝑑 + 𝐼

𝑗ℎ𝑜𝑠𝑝 + (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜

𝑗+ 𝐼𝑖𝑠𝑜

𝑗𝑚𝑖𝑙𝑑 + 𝐼𝑖𝑠𝑜

𝑗ℎ𝑜𝑠𝑝) )

+ (1 − ρ𝑣𝑖𝑛𝑡

)(1 − 𝑉𝐸𝑖𝑛𝑓

)(𝐸2𝑣𝑗 + 𝐼𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑣

𝑗

ℎ𝑜𝑠𝑝+ (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜𝑣

𝑗 + 𝐼𝑖𝑠𝑜𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑖𝑠𝑜𝑣

𝑗

ℎ𝑜𝑠𝑝)))/𝑁

𝑗)

𝑑𝐸1𝑖/𝑑𝑡 = 𝑆

𝑖 β

( Σ𝑗 𝐶

𝑖𝑗 ((1 − ρ

𝑖𝑛𝑡)(𝐸2

𝑗+ 𝐼

𝑗𝑚𝑖𝑙𝑑 + 𝐼

𝑗ℎ𝑜𝑠𝑝 + (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜

𝑗 + 𝐼𝑖𝑠𝑜

𝑗𝑚𝑖𝑙𝑑 + 𝐼𝑖𝑠𝑜

𝑗ℎ𝑜𝑠𝑝) )

+ (1 − ρ𝑣𝑖𝑛𝑡

)(1 − 𝑉𝐸𝑖𝑛𝑓

)(𝐸2𝑣𝑗 + 𝐼𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑣

𝑗

ℎ𝑜𝑠𝑝+ (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜𝑣

𝑗 + 𝐼𝑖𝑠𝑜𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑖𝑠𝑜𝑣

𝑗

ℎ𝑜𝑠𝑝)))/𝑁

𝑗)

− 𝑔1

· 𝐸1𝑖

𝑑𝐸1𝑣𝑖/𝑑𝑡 =

(1 − 𝑉𝐸𝑠𝑢𝑠𝑐

) 𝑆𝑣

𝑖 β( Σ

𝑗 𝐶

𝑖𝑗 ((1 − ρ

𝑖𝑛𝑡)(𝐸2

𝑗+ 𝐼

𝑗𝑚𝑖𝑙𝑑 + 𝐼

𝑗ℎ𝑜𝑠𝑝 + (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜

𝑗+ 𝐼𝑖𝑠𝑜

𝑗𝑚𝑖𝑙𝑑 + 𝐼𝑖𝑠𝑜

𝑗ℎ𝑜𝑠𝑝) )

+ (1 − ρ𝑣𝑖𝑛𝑡

)(1 − 𝑉𝐸𝑖𝑛𝑓

)(𝐸2𝑣𝑗 + 𝐼𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑣

𝑗

ℎ𝑜𝑠𝑝+ (1 − ρ

𝑖𝑠𝑜)(𝐸2𝑖𝑠𝑜𝑣

𝑗 + 𝐼𝑖𝑠𝑜𝑣

𝑗

𝑚𝑖𝑙𝑑+ 𝐼𝑖𝑠𝑜𝑣

𝑗

ℎ𝑜𝑠𝑝)))/𝑁

𝑗)

− 𝑔1

· 𝐸1𝑣𝑖

𝑑𝐸2𝑖/𝑑𝑡 = 𝑔

1· 𝐸1

𝑖 − 𝑔

2· 𝐸2

𝑖 − ν

𝑡𝑒𝑠𝑡,𝑖· 𝐸2

𝑖

𝑑𝐸2𝑣𝑖/𝑑𝑡 = 𝑔

1· 𝐸1𝑣

𝑖 − 𝑔

2· 𝐸2𝑣

𝑖 − 𝑣𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐸2𝑣

𝑖

𝑑𝐸2𝑖𝑠𝑜𝑖/𝑑𝑡 = ν

𝑡𝑒𝑠𝑡,𝑖· 𝐸2

𝑖 − 𝑔

2· 𝐸2𝑖𝑠𝑜

𝑖

𝑑𝐸2𝑖𝑠𝑜𝑣𝑖/𝑑𝑡 = 𝑣𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐸2𝑣

𝑖 − 𝑔

2· 𝐸2𝑖𝑠𝑜𝑣

𝑖

𝑑𝐼𝑖

𝑚𝑖𝑙𝑑/𝑑𝑡 = (1 − 𝑝ℎ𝑜𝑠𝑝𝑖) · 𝑔

2· 𝐸2

𝑖− 𝑔

3· 𝐼

𝑖

𝑚𝑖𝑙𝑑 − 𝑣𝑡𝑒𝑠𝑡,𝑖

· 𝐼𝑖

𝑚𝑖𝑙𝑑

𝑑𝐼𝑣𝑖

𝑚𝑖𝑙𝑑/𝑑𝑡 = (1 − 𝑝ℎ𝑜𝑠𝑝

𝑖· (1 − 𝑉𝐸

𝑠𝑒𝑣)) 𝑔

2· 𝐸2𝑣

𝑖− 𝑔

3· 𝐼𝑣

𝑖

𝑚𝑖𝑙𝑑 − 𝑣𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐼𝑣

𝑖

𝑚𝑖𝑙𝑑

2

𝑑𝐼𝑖𝑠𝑜𝑖

𝑚𝑖𝑙𝑑/𝑑𝑡 = (1 − 𝑝ℎ𝑜𝑠𝑝𝑖) · 𝑔

2· 𝐸2𝑖𝑠𝑜

𝑖+ 𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐼

𝑖

𝑚𝑖𝑙𝑑 − 𝑔3· 𝐼𝑖𝑠𝑜

𝑖

𝑚𝑖𝑙𝑑

𝑑𝐼𝑖𝑠𝑜𝑣𝑖

𝑚𝑖𝑙𝑑/𝑑𝑡 = (1 − 𝑝ℎ𝑜𝑠𝑝

𝑖· (1 − 𝑉𝐸

𝑠𝑒𝑣)) 𝑔

2· 𝐸2𝑖𝑠𝑜𝑣

𝑖+ 𝑣𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐼𝑣

𝑖

𝑚𝑖𝑙𝑑− 𝑔

3· 𝐼𝑖𝑠𝑜𝑣

𝑖

𝑚𝑖𝑙𝑑

𝑑𝑅𝑖

/𝑑𝑡 = 𝑔3· 𝐼

𝑖

𝑚𝑖𝑙𝑑 + 𝑔3· 𝐼𝑖𝑠𝑜

𝑖

𝑚𝑖𝑙𝑑

𝑑𝑅𝑣𝑖

/𝑑𝑡 = 𝑔3

· 𝐼𝑣𝑖

𝑚𝑖𝑙𝑑 + 𝑔

3· 𝐼𝑖𝑠𝑜𝑣

𝑖

𝑚𝑖𝑙𝑑

𝑑𝐼𝑖

ℎ𝑜𝑠𝑝/𝑑𝑡 = 𝑝ℎ𝑜𝑠𝑝𝑖 · 𝑔

2· 𝐸2

𝑖− 𝑔

3· 𝐼

𝑖

ℎ𝑜𝑠𝑝 − 𝑣𝑡𝑒𝑠𝑡,𝑖

· 𝐼𝑖

ℎ𝑜𝑠𝑝

𝑑𝐼𝑣𝑖

ℎ𝑜𝑠𝑝/𝑑𝑡 = 𝑝ℎ𝑜𝑠𝑝

𝑖· (1 − 𝑉𝐸

𝑠𝑒𝑣) · 𝑔

2· 𝐸2𝑣

𝑖− 𝑔

3· 𝐼𝑣

𝑖

ℎ𝑜𝑠𝑝 − 𝑣𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐼𝑣

𝑖

ℎ𝑜𝑠𝑝

𝑑𝐼𝑖𝑠𝑜𝑖

ℎ𝑜𝑠𝑝/𝑑𝑡 = 𝑝ℎ𝑜𝑠𝑝𝑖 · 𝑔

2· 𝐸2𝑖𝑠𝑜

𝑖+ 𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐼

𝑖

ℎ𝑜𝑠𝑝 − 𝑔3

· 𝐼𝑖𝑠𝑜𝑖

ℎ𝑜𝑠𝑝

𝑑𝐼𝑖𝑠𝑜𝑣𝑖

ℎ𝑜𝑠𝑝/𝑑𝑡 = 𝑝ℎ𝑜𝑠𝑝

𝑖· (1 − 𝑉𝐸

𝑠𝑒𝑣) · 𝑔

2· 𝐸2𝑖𝑠𝑜𝑣

𝑖+ 𝑣𝑣

𝑡𝑒𝑠𝑡,𝑖· 𝐼𝑣

𝑖

ℎ𝑜𝑠𝑝− 𝑔

3· 𝐼𝑖𝑠𝑜𝑣

𝑖

ℎ𝑜𝑠𝑝

𝑑Ī𝐻𝑖

/𝑑𝑡 = 𝑔3

· 𝐼𝑖

ℎ𝑜𝑠𝑝 + 𝑔3· 𝐼𝑖𝑠𝑜

𝑖

ℎ𝑜𝑠𝑝 − 𝑔4· Ī𝐻

𝑖

𝑑Ī𝐻𝑣𝑖

/𝑑𝑡 = 𝑔3

· 𝐼𝑣𝑖

ℎ𝑜𝑠𝑝+ 𝑔

3· 𝐼𝑖𝑠𝑜𝑣

𝑖

ℎ𝑜𝑠𝑝− 𝑔

4· Ī𝐻𝑣

𝑖

𝑑𝐻𝑖

/𝑑𝑡 = 𝑔4

· Ī𝐻𝑖

𝑑𝐻𝑣𝑖/𝑑𝑡 = 𝑔

4· Ī𝐻𝑣

𝑖

where we let:

- denote the coefficient of the contact matrix,𝐶𝑖𝑗

, (𝑖, 𝑗) ∈ {1, ..., 13}2

- the superscripts v indicate the different vaccinated compartments,- the subscripts indicate the age groups,𝑖- denote the population size for the age class ,𝑁

𝑗𝑗

- denote the transmission rate,β

3

- denote the rate at which an exposed individual becomes infectious and we set its𝑔1

value days. We set day, and days resulting in an1/𝑔1

= 4 1/𝑔2

= 1 1/𝑔3

= 3

average infectious period of 4 days,- denote the rate of hospital admissions and we set days,1𝑔

41/𝑔

4= 4

- denote the reduction in the transmission rate for unvaccinated individuals (impactρ𝑖𝑛𝑡

of non-pharmaceutical interventions),

- denote the reduction in the transmission rate for vaccinated individuals (impactρ𝑣𝑖𝑛𝑡

of non-pharmaceutical interventions),

- denote the reduction in the transmission rate for isolated individuals,ρ𝑖𝑠𝑜

- denote the rate of testing for unvaccinated individuals.ν𝑡𝑒𝑠𝑡,𝑖

- denote the rate of testing for vaccinated individuals,ν𝑣𝑡𝑒𝑠𝑡,𝑖

- , , and denote the effectiveness of the vaccines on reducing the𝑉𝐸𝑠𝑒𝑣

𝑉𝐸𝑖𝑛𝑓

𝑉𝐸𝑠𝑢𝑠𝑐

probability of hospitalization, the infectiousness and the probability of becominginfected of vaccinated individuals compared to unvaccinated individuals.

Iterative testing

Let assume the proportion of the population of the age class participating in iterative𝑝𝑖𝑡𝑒𝑠𝑡 𝑖

testing. In the scenario where only unvaccinated individuals aged ≥12 y.o. take part initerative testing, we set

andν𝑡𝑒𝑠𝑡,𝑖

= 𝑝𝑖𝑡𝑒𝑠𝑡 · 𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 · (1/𝑡𝑒𝑠𝑡

𝑑𝑒𝑙𝑎𝑦)

ν𝑣𝑡𝑒𝑠𝑡,𝑖

= 0

where denotes the test sensitivity (equal to 75% if the test is self administered𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦and or 90% if it is performed by a professional), and represents the number of days𝑡𝑒𝑠𝑡

𝑑𝑒𝑙𝑎𝑦

between two consecutive tests (7 days in the baseline scenario).In the scenario where individuals participating in the testing campaign are drawn randomly inthe population aged ≥12 y.o. (vaccinated and unvaccinated) we set

,ν𝑡𝑒𝑠𝑡,𝑖

= ν𝑣𝑡𝑒𝑠𝑡,𝑖

= 𝑝𝑖𝑡𝑒𝑠𝑡 · 𝑝

𝑢𝑛𝑣𝑎𝑐𝑐𝑖𝑛𝑎𝑡𝑒𝑑,𝑖· 𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 · (1/𝑡𝑒𝑠𝑡

𝑑𝑒𝑙𝑎𝑦)

where represent the proportion of unvaccinated individuals of the age class .𝑝𝑢𝑛𝑣𝑎𝑐𝑐𝑖𝑛𝑎𝑡𝑒𝑑,𝑖

𝑖

Initialization of the model on September 1st, 2021

On September 1st, 2021, we assume that 25% of the metropolitan French population (range:20%-30%) developed immunity through natural infection. To account for heterogeneity in therisk of infection between the different age groups of the population, we use the distribution ofinfections predicted by a dynamical model calibrated on data until May 15th 20211 (FigureS1). The natural infections are thus distributed across different age groups to reproduce boththe distribution of infections obtained from the model and the proportion of the populationhaving acquired immunity. We also build several scenarios regarding the vaccine coveragesreached in different groups of the population:

4

● 90% or 95% among those older than 60 years old (y.o.)● 60%, 80% or 90% among those aged 18-59 y.o.● 0%, 30% or 70% among the 12-17 y.o.

In our baseline scenario we assume a vaccination coverage of 70%, 80% and 90% among12-17 y.o., 18-59 y.o. and over 60 y.o. respectively on September 1st, 2021.

Figure S1: Fit to the data. Daily hospital admissions through time. The blue line and areacorrespond to model posterior mean and 95% credible intervals, while the grey linecorresponds to data.

5

Additional results

Supplementary Table 1: Relative risk of infection, transmission and hospitalization forunvaccinated individuals relative to vaccinated individuals, in different age groups.This is for our baseline scenario characterized by R0=5 and a vaccine coverage of70%-80%-90% among 12-17 y.o., 18-59 y.o. and over 60 y.o.

Age group Infection Transmission Hospitalisation

0-17 1.3 2.2 18.9

18-59 1.9 3.8 15.0

60+ 2.1 4.3 17.3

All 2.0 4.3 6.1

6

Figure S2: Comparison of the costs of different strategies. Costs of strategies targeting50% of the unvaccinated individuals older than 12 y.o. as a function of the vaccine coveragereached in different groups. The 3 strategies are: weekly testing with an antigenic testperformed by a professional (“antigenic”), weekly testing with an antigenic test performed bythe individual (“autotest”), vaccination.

7

Figure S3: Distribution of infections between groups defined by their age andvaccination status. A. for R0 = 3. B. for R0 = 4. C. for R0 = 5. D. for R0 = 6. The distributionis reported for infections occurring between September 1st, 2021 and March 20th, 2022 (endof the study period) and as a function of the vaccine coverage reached in the 12-17 y.o.,18-59 y.o. and over 60 y.o.

8

Figure S4: Distribution of hospitalisations between groups defined by their age andvaccination status. A. for R0 = 3. B. for R0 = 4. C. for R0 = 5. D. for R0 = 6. The distributionis reported for hospitalizations occurring between September 1st, 2021 and March 20th,2022 (end of the study period) and as a function of the vaccine coverage reached in the12-17 y.o., 18-59 y.o. and over 60 y.o.

9

Figure S5: Contribution of groups defined by their age and vaccination status toinfections, disease spread and hospital burden in a scenario where children aged 0-9y.o. are 50% less infectious than adults, in addition to being 50% less susceptible. Thisis done under our baseline assumptions with R0=5 and a vaccine coverage of70%-80%-90% among 12-17 y.o., 18-59 y.o. and over 60 y.o. Age distribution of newinfections A. in the entire population and B. among vaccinated and unvaccinatedindividuals. Proportion of infections C. attributable to different age groups and D. attributableto different age groups among vaccinated and unvaccinated individuals. Age distribution of

10

hospitalisations E. in the entire population and F. among vaccinated and unvaccinatedindividuals. In all panels, the diamonds indicate the age distribution of the different groups inthe population.

11

Figure S6: Contribution of groups defined by their age and vaccination status toinfections, disease spread and hospital burden in a scenario where the efficacy of thevaccines against infection is set to 80%. This is done under our baseline assumptionswith R0=5 and a vaccine coverage of 70%-80%-90% among 12-17 y.o., 18-59 y.o. and over60 y.o. Age distribution of new infections A. in the entire population and B. amongvaccinated and unvaccinated individuals. Proportion of infections C. attributable to differentage groups and D. attributable to different age groups among vaccinated and unvaccinatedindividuals. Age distribution of hospitalisations E. in the entire population and F. among

12

vaccinated and unvaccinated individuals. In all panels, the diamonds indicate the agedistribution of the different groups in the population.

13

Figure S7: Contribution of groups defined by their age and vaccination status toinfections, disease spread and hospital burden in a scenario where the efficacy of thevaccines against hospitalisation is set to 90%. This is done under our baselineassumptions with R0=5 and a vaccine coverage of 70%-80%-90% among 12-17 y.o., 18-59y.o. and over 60 y.o. Age distribution of new infections A. in the entire population and B.among vaccinated and unvaccinated individuals. Proportion of infections C. attributable todifferent age groups and D. attributable to different age groups among vaccinated and

14

unvaccinated individuals. Age distribution of hospitalisations E. in the entire population andF. among vaccinated and unvaccinated individuals. In all panels, the diamonds indicate theage distribution of the different groups in the population.

15

Figure S8: Projections in the absence of control measures, as a function of the basicreproduction number R0 and vaccine coverage. Peak in daily hospital admissions in theabsence of control measures.

16

Figure S9: Peak in daily hospital admissions under different testing strategies. A. Forself testing (sensitivity: 75%) B. For tests performed by a professional (sensitivity: 90%).The following interventions are explored: Baseline - no intervention; Test every x daysunvaccinated - 50% or 25% of the unvaccinated individuals older than 12 y.o. are testedevery x days; Random - the same number of individuals are tested but in the population ofindividuals older than 12 y.o. irrespective of vaccination status; Vaccinate x% - x% of theunvaccinated individuals older than 12 y.o. are vaccinated. Results are displayed for R0=5.0.We assume 25% of the population has acquired protection through natural infection (range20%-30% corresponding to the vertical bars).

17

1. Salje, H. et al. Estimating the burden of SARS-CoV-2 in France. Science 369, 208–211

(2020).

2. Tran Kiem, C. et al. Short and medium-term challenges for COVID-19 vaccination: from

prioritisation to the relaxation of measures. (2021).

3. Bosetti, P. et al. Impact of mass testing during an epidemic rebound of SARS-CoV-2: a

modelling study using the example of France. Euro Surveill. 26, (2021).

4. Viner, R. M. et al. Susceptibility to SARS-CoV-2 Infection Among Children and

Adolescents Compared With Adults: A Systematic Review and Meta-analysis. JAMA

Pediatr. 175, 143–156 (2021).

5. Davies, N. G. et al. Age-dependent effects in the transmission and control of COVID-19

epidemics. Nature Medicine vol. 26 1205–1211 (2020).

6. FitzJohn, R. odin: ODE Generation and Integration. (2020).

7. Twohig, K. A. et al. Hospital admission and emergency care attendance risk for

SARS-CoV-2 delta (B.1.617.2) compared with alpha (B.1.1.7) variants of concern: a

cohort study. Lancet Infect. Dis. (2021) doi:10.1016/S1473-3099(21)00475-8.

8. Bager, P. et al. Increased Risk of Hospitalisation Associated with Infection with

SARS-CoV-2 Lineage B.1.1.7 in Denmark. SSRN Electronic Journal

doi:10.2139/ssrn.3792894.

9. Lapidus, N. et al. Do not neglect SARS-CoV-2 hospitalization and fatality risks in the

middle-aged adult population. Infectious Diseases Now vol. 51 380–382 (2021).

18